VisPlay: Self-Evolving Vision-Language Models from Images

TL;DR

VisPlay introduces a self-evolving reinforcement learning framework that enables vision-language models to autonomously improve their reasoning abilities using unlabeled image data, reducing reliance on human annotations.

Contribution

It proposes a novel RL framework with dual roles and a new training algorithm, scalable across different model families, for autonomous improvement of VLMs.

Findings

Achieves consistent improvements in visual reasoning and generalization.

Reduces hallucinations in model outputs.

Scales effectively across multiple model architectures.

Abstract

Reinforcement learning (RL) provides a principled framework for improving Vision-Language Models (VLMs) on complex reasoning tasks. However, existing RL approaches often rely on human-annotated labels or task-specific heuristics to define verifiable rewards, both of which are costly and difficult to scale. We introduce VisPlay, a self-evolving RL framework that enables VLMs to autonomously improve their reasoning abilities using large amounts of unlabeled image data. Starting from a single base VLM, VisPlay assigns the model into two interacting roles: an Image-Conditioned Questioner that formulates challenging yet answerable visual questions, and a Multimodal Reasoner that generates silver responses. These roles are jointly trained with Group Relative Policy Optimization (GRPO), which incorporates diversity and difficulty rewards to balance the complexity of generated questions with the quality of the silver answers. VisPlay scales efficiently across two model families. When trained on Qwen2.5-VL and MiMo-VL, VisPlay achieves consistent improvements in visual reasoning, compositional generalization, and hallucination reduction across eight benchmarks, including MM-Vet and MMMU, demonstrating a scalable path toward self-evolving multimodal intelligence. The project page is available at https://bruno686.github.io/VisPlay/